Y
RAMON ALAMÜS
AIRBORNE SENSOR INTEGRATION AND DIRECT ORIENTATION OF THE CASI SYSTEM
Ramon ALAMÜS, Julià TALAYA
Institut Cartografic de Catalunya
Parc de Montjuïc 08038 Barcelona Spain
ralamus @icc.es, talaya@icc.es
KEY WORDS: Sensor integration, Self-calibration, INS/GPS integration
ABSTRACT
The ICC (Institut Cartografic de Catalunya) has developed a SISA (Sistema Integrat de Sensors Aerotransportats - Inte-
grated System of Airborne Sensors). SISA is an integrated system of airborne sensors (imagery, position, attitude,...) and
algorithms that can be used to obtain correct georeferentiation of imaging sensors. Thus far SISA has been used with the
CASI (Compact Airborne Spectographic Imager) sensor and it has also been used in one gravimetric flight.
The main goal of SISA is to provide precise attitude and positioning to an imaging sensor. To carry out this task, SISA
integrates GPS and IMU (Inertial Measurement Unit) data and provides a synchronization procedure for the entire set of
sensors. The current configuration of SISA provides an interface to the attitude subsystem (based on a Litton LTN 101
FLAGSHIP INS - Inertial Navigation System), an interface to a dual frequency GPS receiver and a robust procedure for
synchronizing the attitude sensor (IMU/INS) and the imaging sensor (CASI). ICC has developed software to ensure the
correct time tag of the inertial and image data and to determine the trajectory and attitude of the airborne platform.
The discrepancies between the Image Reference System and the Inertial Reference System due to the mounting of the
sensors are determined together with certain calibration parameters of the CASI in a bundle adjustment using the Geo-
TeX/ACX software.
Direct methods for orientation require a good knowledge (a priori) of the geometric relationships between the sensors
involved. This paper discusses the stability of the SISA-CASI attachment, comparing adjusted misalignment matrix and
self-calibration parameters in series of CASI flights.
As a conclusion, SISA proves its capabilities for the absolute direct orientation of the CASI imaging sensor.
1 INTRODUCTION
Linear sensors like the CASI system have a very weak geometry because each line has a different set of orientation pa-
rameters (CASI images usually contain 5000 to 7000 lines). For practical reasons the orientation should be determined
using direct methods that rely on the use of GPS and inertial systems.
For mapping purposes, it is necessary to attain orientation accuracies below 1 - 2 pixels for allowing further mosaicking
and image fusion.
The accuracy required on the position depends on the pixel size. In CASI projects it ranges from 2.5 m to 10 m, while
attitude precision depends on the instantaneous field of view (IFOV) of a single pixel, which is 5 arcminutes in CASI’s
current configuration.
2 CASI SYSTEM
The imaging subsystem (the CASI sensor) is a pushbroom linear scanner based on a matrix CCD, which allows collection
of up to 19 spectral bands, selected from 288 spectral samples in the range of 430 nm to 950 nm and 512 spatial samples.
Initially the orientation of the CASI relied on a GPS receiver for positioning and on SPERRY VG-14A dual axis gyroscope
for pitch and roll corrections. The original orientation approach was not accurate enough to allow mosaicking CASI
images or general fusion with orthoimages (Colomina 95b). Root mean squares of residuals of roll and pitch are 0.24°
and 0.13° respectively (i.e. 2.75 pixels and 1.5 pixels) in dynamic mode. The lack of a heading sensor was a major
handicap because it was not possible to recover a posteriori all the variations in crab, even though it was possible to
compute a mean crab per strip in a bundle adjustment.
The synchronization procedure was not robust. Synchronization errors lie in a wrong time tag for each single line of the
CASI causing that the orientation of the airplane is assigned to a linear image that was taken at a different instant.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B1. Amsterdam 2000. 5